A novel auto-pruned ensemble clustering via SOCP

Abstract

Operations Research (OR) plays a crucial role in strategic decision-making in today's business world; it uses complex algorithms and data analytic to provide decision-makers with the necessary information. The proposed study presents a novel ensemble clustering for decision making in various disciplines including OR problems which introduce a second-order conic optimization model. This method provides a significant advantage over the traditional difference of convex programming by continuously and convexly solving integer programming. The model optimizes the balance between accuracy and diversity, resulting in the selection of the best candidates for prediction. The study's remarkable contribution lies in the automatic sub-ensemble selection while optimizing for accuracy and diversity. The model has been verified using real data and achieves competitive prediction performance. Furthermore, this approach illustrates how OR can be utilized to enhance ensemble clustering and decision-making.